TY - JOUR
T1 - Ecological validity of a deep learning algorithm to detect gait events from real-life walking bouts in mobility-limiting diseases
AU - Romijnders, Robbin
AU - Salis, Francesca
AU - Hansen, Clint
AU - Küderle, Arne
AU - Paraschiv-Ionescu, Anisoara
AU - Cereatti, Andrea
AU - Alcock, Lisa
AU - Aminian, Kamiar
AU - Becker, Clemens
AU - Bertuletti, Stefano
AU - Bonci, Tecla
AU - Brown, Philip
AU - Buckley, Ellen
AU - Cantu, Alma
AU - Carsin, Anne Elie
AU - Caruso, Marco
AU - Caulfield, Brian
AU - Chiari, Lorenzo
AU - D'Ascanio, Ilaria
AU - Del Din, Silvia
AU - Eskofier, Björn
AU - Fernstad, Sara Johansson
AU - Fröhlich, Marceli Stanislaw
AU - Garcia Aymerich, Judith
AU - Gazit, Eran
AU - Hausdorff, Jeffrey M.
AU - Hiden, Hugo
AU - Hume, Emily
AU - Keogh, Alison
AU - Kirk, Cameron
AU - Kluge, Felix
AU - Koch, Sarah
AU - Mazzà, Claudia
AU - Megaritis, Dimitrios
AU - Micó-Amigo, Encarna
AU - Müller, Arne
AU - Palmerini, Luca
AU - Rochester, Lynn
AU - Schwickert, Lars
AU - Scott, Kirsty
AU - Sharrack, Basil
AU - Singleton, David
AU - Soltani, Abolfazl
AU - Ullrich, Martin
AU - Vereijken, Beatrix
AU - Vogiatzis, Ioannis
AU - Yarnall, Alison
AU - Schmidt, Gerhard
AU - Maetzler, Walter
N1 - Publisher Copyright:
Copyright © 2023 Romijnders, Salis, Hansen, Küderle, Paraschiv-Ionescu, Cereatti, Alcock, Aminian, Becker, Bertuletti, Bonci, Brown, Buckley, Cantu, Carsin, Caruso, Caulfield, Chiari, D'Ascanio, Del Din, Eskofier, Fernstad, Fröhlich, Garcia Aymerich, Gazit, Hausdorff, Hiden, Hume, Keogh, Kirk, Kluge, Koch, Mazzà, Megaritis, Micó-Amigo, Müller, Palmerini, Rochester, Schwickert, Scott, Sharrack, Singleton, Soltani, Ullrich, Vereijken, Vogiatzis, Yarnall, Schmidt and Maetzler.
PY - 2023
Y1 - 2023
N2 - Introduction: The clinical assessment of mobility, and walking specifically, is still mainly based on functional tests that lack ecological validity. Thanks to inertial measurement units (IMUs), gait analysis is shifting to unsupervised monitoring in naturalistic and unconstrained settings. However, the extraction of clinically relevant gait parameters from IMU data often depends on heuristics-based algorithms that rely on empirically determined thresholds. These were mainly validated on small cohorts in supervised settings. Methods: Here, a deep learning (DL) algorithm was developed and validated for gait event detection in a heterogeneous population of different mobility-limiting disease cohorts and a cohort of healthy adults. Participants wore pressure insoles and IMUs on both feet for 2.5 h in their habitual environment. The raw accelerometer and gyroscope data from both feet were used as input to a deep convolutional neural network, while reference timings for gait events were based on the combined IMU and pressure insoles data. Results and discussion: The results showed a high-detection performance for initial contacts (ICs) (recall: 98%, precision: 96%) and final contacts (FCs) (recall: 99%, precision: 94%) and a maximum median time error of −0.02 s for ICs and 0.03 s for FCs. Subsequently derived temporal gait parameters were in good agreement with a pressure insoles-based reference with a maximum mean difference of 0.07, −0.07, and <0.01 s for stance, swing, and stride time, respectively. Thus, the DL algorithm is considered successful in detecting gait events in ecologically valid environments across different mobility-limiting diseases.
AB - Introduction: The clinical assessment of mobility, and walking specifically, is still mainly based on functional tests that lack ecological validity. Thanks to inertial measurement units (IMUs), gait analysis is shifting to unsupervised monitoring in naturalistic and unconstrained settings. However, the extraction of clinically relevant gait parameters from IMU data often depends on heuristics-based algorithms that rely on empirically determined thresholds. These were mainly validated on small cohorts in supervised settings. Methods: Here, a deep learning (DL) algorithm was developed and validated for gait event detection in a heterogeneous population of different mobility-limiting disease cohorts and a cohort of healthy adults. Participants wore pressure insoles and IMUs on both feet for 2.5 h in their habitual environment. The raw accelerometer and gyroscope data from both feet were used as input to a deep convolutional neural network, while reference timings for gait events were based on the combined IMU and pressure insoles data. Results and discussion: The results showed a high-detection performance for initial contacts (ICs) (recall: 98%, precision: 96%) and final contacts (FCs) (recall: 99%, precision: 94%) and a maximum median time error of −0.02 s for ICs and 0.03 s for FCs. Subsequently derived temporal gait parameters were in good agreement with a pressure insoles-based reference with a maximum mean difference of 0.07, −0.07, and <0.01 s for stance, swing, and stride time, respectively. Thus, the DL algorithm is considered successful in detecting gait events in ecologically valid environments across different mobility-limiting diseases.
KW - deep learning (artificial intelligence)
KW - free-living
KW - gait analysis
KW - gait events detection
KW - inertial measurement unit (IMU)
KW - mobility
UR - http://www.scopus.com/inward/record.url?scp=85175557037&partnerID=8YFLogxK
U2 - 10.3389/fneur.2023.1247532
DO - 10.3389/fneur.2023.1247532
M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???
C2 - 37909030
AN - SCOPUS:85175557037
SN - 1664-2295
VL - 14
JO - Frontiers in Neurology
JF - Frontiers in Neurology
M1 - 1247532
ER -